Subband Simulation in MATLAB

VLADISLAV SKORPIL, ABDULHAKIM ABUZAHU

Department of Telekommunications,
Brno University of Technology,
Purkynova 118, 612 00 Brno,
CZECH REPUBLIC,
http://www.vutbr.cz/


Abstract: - In the start of the flow chart, there are two steps, which are prepared for new input samples.
Firstly, all the samples in the buffer (512 samples) are down shifted and then put new 32 samples at the end of
the buffer, which is required for further calculations. But in this work different algorithm, which contain both
in only one step is used. One more variable known as ofs is present, which shows where is the end of the
previous data read, because 32 samples into 512 samples buffer is read, it is different number every time.

Key-Words: - Subband, sample, buffer, algorithm



1 Introduction

Instead of shifting in each run of flowchart (it is 12
times in one frame) only offset is shifted. It gives
more time for other algorithms. When 32 new
samples are added, it gives 512 together, which is
all defined length for buffer for windowing. In
next step is windowing with coefficients take from
standard. Figure 1 shows analysis of filter bank.
These are coefficients for matrixing in previous
flowchart. These coefficients are calculated by the
following formula:

( )( )
1
]
1

+
64
16 1 2 cos

k i Mik (1)


2 Evaluation of Signal Quality

Signal quality can be evaluated using various
criteria; it can essentially divide these into
subjective and objective criteria. Subjective
parameters of an acoustic signal are determined on
the basis of the psycho acoustic properties of the
hearing equipment. The objective parameters of
the signal are the physical parameters. These
include parameters in the temporal domain (middle
value, variance, slope and peak coefficients,
correlation functions, etc.) and the frequency

Fig.1 Filter bank analysis

domain (spectrum or the spectral output density of
the signal). Research on the subjective quality of
acoustic signals both speech and music has
received a lot of attention, as no objective criterion
capable of fully replacing subjective listening tests
exists at present, a detailed methodology for
performing listening tests exists under e.g. ITU-R..
In this work, I use, for listening tests, an evaluation
scale based on the recommendations of ITU-R; see
Tab. 1. Since listening tests place high demands on
equipment that I cannot meet, and since I cannot
perform tests on large groups of people and then
statistically process them, my results may,
however, differ from those acquired precisely
according to recommendations, and are for
information only. The median quadratic difference
between the original signal and the compressed
signal is an important objective criterion for
judging signal quality. This criterion can in some
cases also convey the subjective quality of a
signal. In the case of the MPEG codec,
psychoacoustic model guarantees (within the limits
of its precision) that distortion caused by
compression will be as inaudible as possible and,
to some degree, will be remedied by the listener.


Table 1 Five Degree scale fro the worsening
of Signal Quality (ITU-R)


3 Scale factor Calculating

A scale technique is used in the MPEG-1 coding
scheme, which provides an effective overall
dynamic range of 120 dB/subband with a
resolution of 2dB/scale factor class. The
calculation of the scale factors is made for every
12 subband samples. The maximum value of these
samples is determined. The lowest value, which is
larger than this maximum value is used as scale-
factor. Scale-factor can be calculated after
psychoacoustic model according to standard. In
fact, in MATLAB program, the scale-factor was
chosen before psychoacoustic model. The
algorithm contains to looking for maximum value
of samples and after searching minimum
appropriate scale-factor.
The calculated scale-factor is presented by 6 bits
word. It is transmitted only if a non-zero number
of bits have been allocated to the subband. Coding
of scale-factors is presented in MATLAB. When
minimal scale-factor is find out, then information
about it is stored in scfi(i) variable. The data about
scale-factors are then given to decoders through
this global vector of scale-factor indexes.


4 FFT Analysis

The masking threshold is derived from an
estimate of the power density spectrum that is
calculated by a 512-point FFT. The FFT is
calculated directly from the input PCM signal,
windowed by a Hann window (Equation 2)
displayed in Figure 2.

1 0
2
cos 1 5 . 0
3
8
) (

,
_

1
]
1

N i
N
i
i h

(2)
For a coincidence in time between the bit
allocation and the corresponding subband
samples, the PCM samples entering the FFT
have to be displayed. The delay of the analysis
subband filter in 256 samples, corresponding
to 5.3 ms at the 48 kHz sampling rate. A
window shift of 256 samples is required to
compensate for the delay in the analysis
subband filter. In addition, the Hann window
must coincide with the subband samples of the
frame. For Layer-1, this amount to an
additional window shift of 64 samples.
When computing is at the end of 512 energy
samples, values are reindexed. It means that
calculated energy is shifting (by indexing) in
corresponding position. New indexes are taken
by butterfly algorithm as shown in Figure 3. It
introduced 8-point butterfly for taking
computing principle of FFT.



Fig. 2 Hann window in the beginning in
computing of FFT


Fig. 3 Flow graph of-point FFT computation


Computation of FFT in MATLAB program
can be calculated by only one instruction
(procedure), but principle of computing is not
possible to see. Here we use just
multiplication, addition and subtraction in
MATLAB.


Fig.4 Test signal inside buffer after
windowing by Hann window

This part has several loops. Shift of 64 samples
is recommended by standard to keep
coincidence in time, but before buffer is filled
by 384 new samples. For correct mathematical
calculations, it is necessary to prepare 512
samples large buffer for FFT. The rest of
buffer is filled by zeros. When new samples
are put in buffer, is used addressing with
offset. This principle is similar as in first step
of encoding. After windowing by Hann
window as shown in Figure 4, variables are
initialized on zeros values. Thus everything is
ready for starting of calculating FFT output
samples, keeping energy value in each
subband. It is 512 point FFT, it gives 256
values to 32 subbands.




Fig. 5 Energy power calculated in first run

If input signal is applied, MATLAB gives
energy spectrum picture as presented in Figure
5. In first run 384 samples are taken, but FFT
buffer have positions for 512 equal to the size
of FFT. The rest of free positions will be filled
by zeros. And after second, third, fourth, and
other runs buffer will be full. After first run,
the signal will be still the same, so energy in
the next runs will remain the same as shown in
(Figure 6). This shows spread of power in
frequency spectrum.
This is good position in program to start
analysis of the input signal. User can see how
does input signal looks like in the frequency
domain. But for MPEG algorithm, this step is
the most important, because further
computations are based on the FFT output
(energy).


Fig. 6 Energy power calculated after sigth run



5 MATLAB Computing of the
Sound Pressure Level

The more powerful version was chosen for the
determination of the sound pressure level in
program in MATLAB. A maximum is found
from all spectral lines in each subband and one
pressure level gives by scalefactors. In fact, it
means two loops, first is making a searching
through spectral lines in each subband, second
is searching just through subbands comparing
to an appropriate pressure level and maximum
from first loop.
In this section of program, only one loop is
possible to see. It is just first step from both.
second is comparing at the end of the
psychoacoustic model, which is given together
with calculating SNR (signal-to-noise ratio). It
is more powerful as one loop is spared.
The tonality of a masking component has an
influence on the masking threshold. For this
reason, it is worthwhile to discriminate
between tonal and non-tonal components. For
calculating the global masking threshold, it is
necessary to derive the tonal and non-tonal
components from the FFT.
So this step was divided into two parts. First is
aimed to find tones and second to find noise
with high power, which is necessary to
transmit. In fact, it is named noise, but it
should be many tones closer together. In
addition, there is one more part called labeling
of local maxima, but it is concluded in the part
for tones.
This step of standard starts with labeling of
local maxima. It then extracts tonal
components (sinusoids) and calculates the
intensity of the non-tonal components within a
bandwidth of a critical band. The bandwidth of
the critical band varies with the frequency with
a bandwidth of about 0.1 kHz at low and with
a bandwidth of about 4 kHz at high
frequencies. It is known from psychoacoustic
experiments that the ear has a better frequency
resolution in the lower frequency region than
in the higher ones [3]. To determine if a local
maximum is examined, the frequency range df
is given by standard for all sampling
frequencies of MPEG-1.

6 Conclusion

For further information about df for different
frequencies and different layers is better to
study the Reccomendationn just one possibility
for Layer-1 and 44.1 kHz as sampling
frequency. Choosing it in the beginning of this
work causes this aim. Moreover, sampling
frequency 44.1 kHz is the most popular
throughout the world. The second sampling
frequency used very often for digital audio is
48 kHz. First is based mostly on commercial
using and second on professional devices
means studios.

References:
[1] ISO/IEC 13818-7 Information Technology
Generic coding of moving pictures and
Associated Audio Information part 7:
Advanced Audio Coding (AAC), 1996, 131s.
[2] BOSI M. and RANDENBURG, K.
ISO/IEC MPEG-2, Advanced Audio Coding.
Journal of the Audio Engineering Society,
Vol.45, 1997, No. 10,PP. 789-814
[3] SERANTES, C. PENA A., Prelcic N.
A Fast Noise-scaling Algorithm for Uniform
Quantization in Audio Coding schemes. IEEE
Proceedings, 1997 , PP 339-342
[4].ABUZAHU,A. Digital Audio
Compression. Ph.D. Thesis, BUT, Brno 2003

Acknowledgement:
This research was supported by the grants:
No 102/03/0434 Limits for broad-band signal
transmission on the twisted pairs and other
system co-existence. The Grant Agency of the
Czech Republic (GACR)
No 102/03/0260 Development of network
communication application programming
interface for new generation of mobile and
wireless terminals. The Grant Agency of the
Czech Republic (GACR)
No 102/03/0560 New methods for location and
verification of compliance of quality of service
in new generation networks. The Grant
Agency of the Czech Republic (GACR)
No MS 1850022 Research of
communication systems and technologies
(Research design)
Grant 2811 F1 Advanced Technology of
Transport Networks in Education (grant
of the Czech Ministry of Education, Youth
and Sports)
Grant 3112 F1 Inovation of Education of Last
Mile Data Transmission (grant of the Czech
Ministry of Education, Youth and Sports)